Document Type

Conference Paper

Publication Date

2017-05-01

Originally Published In

Proc. Smart Data Pricing (SDP) Workshop 2017

Embargo Period

3-26-2017

Abstract

We consider a cloud service based on spot instances and explore bidding and pricing strategies aimed at optimizing users' utility and provider's revenue, respectively. Our focus is on jobs that are heterogeneous in both valuation and sensitivity to execution delay. Of particular interest is the impact of correlation in these two dimensions. We characterize optimal bidding and pricing strategies under some simplifying assumptions, and more importantly highlight the impact of correlation in determining the benefits of a spot service over an on-demand service. We also provide a preliminary assessment of the results' robustness under more general assumptions.

Share

COinS